Practical AI
I've spent years building dashboards in Tableau. Like many developers, I've invested countless hours learning not just the software itself, but the principles behind good dashboard design. Understanding the audience, knowing which questions need to be answered, and presenting information in a way that helps people make better decisions has always been the part of the job I've enjoyed the most. In fact, it’s one of the things I love most about our philosophy here at Allegro Analytics.
We pride ourselves on being the human side of analytics – because successful analytics has never been about the dashboards themselves. It's about understanding the people using them, the decisions they're trying to make, and the business problems they're trying to solve.
Recently, one of our clients shared that they were planning to take a different approach. Instead of using Tableau like they had done for years, they wanted to use AI to generate dashboards directly from their curated data sources.
I'll be honest, my first thought wasn't, "This is exciting." It was, "What does this mean for someone like me?"
After spending years becoming an expert in Tableau, it was impossible not to wonder where AI fit into the picture. Was it replacing what I do, or was it simply changing how I do it?
The more I thought about it, the more I realized I wasn't going to answer that question by reading articles or watching demos. I needed to get my hands dirty and see what it was actually like to build a dashboard this way.
What surprised me most was how much of the work hadn't changed.
The dashboard might be built with AI, but it still starts with someone understanding the business problem. AI doesn't know what your executives care about or which KPIs are the ones that actually drive decisions. It doesn't know that one department always wants to compare this quarter against last year, or that another team cares more about trends than individual numbers.
Someone still has to define the goal.
As I experimented with creating similar dashboards, I found myself asking exactly the same questions I ask before I ever open up Tableau.
Who is this dashboard for?
What decisions are they trying to make?
Which metrics matter, and just as importantly, which ones don't?
What should someone notice within the first few seconds of opening the dashboard?
Those questions haven't gone away. If anything, they've become more important because AI can only work with the context we give it.
Another realization came pretty quickly: the quality of the data matters even more than it did before. Our client wasn't pointing AI at a collection of raw tables and hoping for the best. They had invested time in building a curated, trusted data source with consistent business definitions and governed calculations. That work turned out to be the foundation that made AI useful in the first place.
As Tableau developers, we've always talked about the importance of a single source of truth. AI doesn't eliminate that need. It amplifies it. If the underlying data isn't organized and well defined, AI will happily build beautiful dashboards that answer the wrong question or use inconsistent metrics.
That isn't an AI problem. It's a data problem.
I also noticed something interesting as I compared AI-generated dashboards to the ones I would normally build myself. The dashboards weren't necessarily wrong. In fact, many of them looked impressive at first glance. But after looking a little closer, I'd often find myself making small adjustments.
Maybe I would choose a different chart because it told the story more clearly. Maybe I'd remove a visualization that didn't add much value or reorganize the layout, so the most important information appeared first. Sometimes the dashboard needed less information, not more.
Those are decisions that come from experience.
They're the kind of choices you make after years of sitting with business users, watching how they interact with dashboards, and learning what actually helps people make decisions instead of simply displaying data.
I also discovered that prompting AI isn't really about finding some magical combination of words. It reminded me much more of gathering requirements for a project.
The more specific I became about the audience, the business goals, the KPIs, and the type of story I wanted the dashboard to tell, the better the results became. A vague request produced a generic dashboard. A well-defined prompt produced something much closer to what I would have designed myself.
That was probably the biggest lesson for me.
The technology may be different, but the skills that make someone a good analytics professional haven't really changed.
Understanding the business.
Knowing the data.
Recognizing when a visualization is misleading.
Asking better questions than anyone else in the room.
Those skills are still incredibly valuable. If anything, AI shifts more of our value away from the mechanics of building dashboards and toward the thinking that happens before and after the dashboard exists.
Looking back, I realize my initial concern wasn't really about Tableau. It was about change. Like a lot of people who have built careers around a particular technology, I wondered whether AI was going to make years of experience irrelevant. Now I don't see it that way.
I still believe Tableau is an incredible platform, and I expect to keep using it. But I also think AI is becoming another tool that belongs in our toolkit. The organizations that get the most value from it won't be the ones replacing analytics professionals; they'll be the ones combining AI with people who understand their business and their data.
This experience reminded me that my career was never really about Tableau. Tableau was simply the tool I used to solve problems. The real expertise has always been understanding data well enough to help people make better decisions. That part hasn't changed. And I don't think it's going away anytime soon.
– Kim Unger
